Search papers, labs, and topics across Lattice.
This paper introduces a Weighted Quality Index (QI) for Automated Program Repair (APR) that extends beyond simple test suite pass rates to encompass maintainability, security, and generation efficiency, inspired by the ISO/IEC 25010 software quality model. Evaluating dense Qwen2.5-Coder and DeepSeek-Coder-V2-Lite MoE models on QuixBugs and Defects4J datasets reveals that model rankings shift significantly based on QI weighting schemes, highlighting the limitations of single-metric evaluation. Notably, the MoE model achieves comparable functional correctness to larger dense models with substantially fewer active parameters, suggesting active parameter count is a more relevant metric for sparse code models.
Sparse MoE models match dense models' code repair correctness with drastically fewer active parameters, challenging the sole reliance on total parameter count for evaluating code generation efficiency.
Automated Program Repair (APR) with language models is usually evaluated by whether a generated patch passes the test suite, which can hide differences in maintainability, security, and computational cost. We propose a Weighted Quality Index (QI), inspired by the ISO/IEC 25010 software quality model, that combines functional correctness, maintainability, security, and generation efficiency under configurable weighting schemes. We evaluate three dense Qwen2.5-Coder models (3B, 7B, 14B) and the 16B-parameter DeepSeek-Coder-V2-Lite Mixture-of-Experts (MoE) model (2.4B active parameters) on 40 QuixBugs and 90 Defects4J bugs, all run locally on identical hardware to control for infrastructure effects. Model rankings change with the weighting scheme, showing that single-metric evaluation can hide trade-offs. The MoE model shows almost no statistically significant difference in correctness from the 7B and 14B dense models (McNemar's exact test) while using 3-6 times fewer active parameters, whereas correctness increases significantly across the three dense scales. These results suggest that active parameter count can be a more informative lens than total parameter count for sparse code models.
